* [PATCH v1 0/4] Enable yellow meter hierarchy
@ 2022-05-13 7:33 Shun Hao
2022-05-13 7:33 ` [PATCH v1 1/4] net/mlx5: support previous meter color aware Shun Hao
` (4 more replies)
0 siblings, 5 replies; 6+ messages in thread
From: Shun Hao @ 2022-05-13 7:33 UTC (permalink / raw)
To: viacheslavo, matan, orika; +Cc: dev, rasland
To support yellow meter hierarchy, need to support meter action in
yellow policy flow, so both green and yellow packet can go to next
meter. Currently there's the limitation that only the same next meter
supported for green and yellow.
Meanwhile, color aware mode is supported, so the next meter can be aware
of the previous color, so can have different process strategy.
Shun Hao (4):
net/mlx5: support previous meter color aware
net/mlx5: support yellow meter action in hierarchy
net/mlx5: support yellow meter action for hierarchy tag rule
net/mlx5: add validation for yellow meter action
doc/guides/nics/mlx5.rst | 6 +-
drivers/net/mlx5/mlx5.c | 8 +-
drivers/net/mlx5/mlx5.h | 10 +-
drivers/net/mlx5/mlx5_flow_aso.c | 19 +-
drivers/net/mlx5/mlx5_flow_dv.c | 406 +++++++++++++++++++----------
drivers/net/mlx5/mlx5_flow_meter.c | 60 +++--
6 files changed, 345 insertions(+), 164 deletions(-)
--
2.20.0
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH v1 1/4] net/mlx5: support previous meter color aware
2022-05-13 7:33 [PATCH v1 0/4] Enable yellow meter hierarchy Shun Hao
@ 2022-05-13 7:33 ` Shun Hao
2022-05-13 7:33 ` [PATCH v1 2/4] net/mlx5: support yellow meter action in hierarchy Shun Hao
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Shun Hao @ 2022-05-13 7:33 UTC (permalink / raw)
To: viacheslavo, matan, orika; +Cc: dev, rasland
This patch adds the support for previous color aware for meter.
Start_color setting is set to UNDEFINED when creating meter object that
is color aware.
Signed-off-by: Shun Hao <shunh@nvidia.com>
Acked-by: Matan Azard <matan@nvidia.com>
---
drivers/net/mlx5/mlx5.h | 2 ++
drivers/net/mlx5/mlx5_flow_aso.c | 19 +++++++++++--------
drivers/net/mlx5/mlx5_flow_meter.c | 3 ++-
3 files changed, 15 insertions(+), 9 deletions(-)
diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
index 23a28f6e52..c92ce6b527 100644
--- a/drivers/net/mlx5/mlx5.h
+++ b/drivers/net/mlx5/mlx5.h
@@ -859,6 +859,8 @@ struct mlx5_flow_meter_info {
uint32_t transfer:1;
uint32_t def_policy:1;
/* Meter points to default policy. */
+ uint32_t color_aware:1;
+ /* Meter is color aware mode. */
void *drop_rule[MLX5_MTR_DOMAIN_MAX];
/* Meter drop rule in drop table. */
uint32_t drop_cnt;
diff --git a/drivers/net/mlx5/mlx5_flow_aso.c b/drivers/net/mlx5/mlx5_flow_aso.c
index eb7fc43da3..4129e3a9e0 100644
--- a/drivers/net/mlx5/mlx5_flow_aso.c
+++ b/drivers/net/mlx5/mlx5_flow_aso.c
@@ -652,6 +652,7 @@ mlx5_aso_mtr_sq_enqueue_single(struct mlx5_dev_ctx_shared *sh,
uint16_t res;
uint32_t dseg_idx = 0;
struct mlx5_aso_mtr_pool *pool = NULL;
+ uint32_t param_le;
rte_spinlock_lock(&sq->sqsl);
res = size - (uint16_t)(sq->head - sq->tail);
@@ -688,15 +689,14 @@ mlx5_aso_mtr_sq_enqueue_single(struct mlx5_dev_ctx_shared *sh,
wqe->aso_dseg.mtrs[dseg_idx].ebs_eir = 0;
}
fmp = fm->profile;
- if (fmp->profile.packet_mode)
- wqe->aso_dseg.mtrs[dseg_idx].v_bo_sc_bbog_mm =
- RTE_BE32((1 << ASO_DSEG_VALID_OFFSET) |
- (MLX5_FLOW_COLOR_GREEN << ASO_DSEG_SC_OFFSET) |
- (MLX5_METER_MODE_PKT << ASO_DSEG_MTR_MODE));
+ param_le = (1 << ASO_DSEG_VALID_OFFSET);
+ if (fm->color_aware)
+ param_le |= (MLX5_FLOW_COLOR_UNDEFINED << ASO_DSEG_SC_OFFSET);
else
- wqe->aso_dseg.mtrs[dseg_idx].v_bo_sc_bbog_mm =
- RTE_BE32((1 << ASO_DSEG_VALID_OFFSET) |
- (MLX5_FLOW_COLOR_GREEN << ASO_DSEG_SC_OFFSET));
+ param_le |= (MLX5_FLOW_COLOR_GREEN << ASO_DSEG_SC_OFFSET);
+ if (fmp->profile.packet_mode)
+ param_le |= (MLX5_METER_MODE_PKT << ASO_DSEG_MTR_MODE);
+ wqe->aso_dseg.mtrs[dseg_idx].v_bo_sc_bbog_mm = RTE_BE32(param_le);
switch (fmp->profile.alg) {
case RTE_MTR_SRTCM_RFC2697:
/* Only needed for RFC2697. */
@@ -709,6 +709,9 @@ mlx5_aso_mtr_sq_enqueue_single(struct mlx5_dev_ctx_shared *sh,
RTE_BE32(1 << ASO_DSEG_BBOG_OFFSET);
break;
case RTE_MTR_TRTCM_RFC4115:
+ wqe->aso_dseg.mtrs[dseg_idx].v_bo_sc_bbog_mm |=
+ RTE_BE32(1 << ASO_DSEG_BO_OFFSET);
+ break;
default:
break;
}
diff --git a/drivers/net/mlx5/mlx5_flow_meter.c b/drivers/net/mlx5/mlx5_flow_meter.c
index a3d1f2c08d..b52de70afa 100644
--- a/drivers/net/mlx5/mlx5_flow_meter.c
+++ b/drivers/net/mlx5/mlx5_flow_meter.c
@@ -1005,7 +1005,7 @@ mlx5_flow_meter_validate(struct mlx5_priv *priv, uint32_t meter_id,
RTE_MTR_ERROR_TYPE_MTR_PARAMS,
NULL, "Meter object params null.");
/* Previous meter color is not supported. */
- if (params->use_prev_mtr_color)
+ if (params->use_prev_mtr_color && !priv->sh->meter_aso_en)
return -rte_mtr_error_set(error, ENOTSUP,
RTE_MTR_ERROR_TYPE_MTR_PARAMS,
NULL,
@@ -1293,6 +1293,7 @@ mlx5_flow_meter_create(struct rte_eth_dev *dev, uint32_t meter_id,
fm->active_state = 1; /* Config meter starts as active. */
fm->is_enable = params->meter_enable;
fm->shared = !!shared;
+ fm->color_aware = !!params->use_prev_mtr_color;
__atomic_add_fetch(&fm->profile->ref_cnt, 1, __ATOMIC_RELAXED);
if (params->meter_policy_id == priv->sh->mtrmng->def_policy_id) {
fm->def_policy = 1;
--
2.20.0
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH v1 2/4] net/mlx5: support yellow meter action in hierarchy
2022-05-13 7:33 [PATCH v1 0/4] Enable yellow meter hierarchy Shun Hao
2022-05-13 7:33 ` [PATCH v1 1/4] net/mlx5: support previous meter color aware Shun Hao
@ 2022-05-13 7:33 ` Shun Hao
2022-05-13 7:33 ` [PATCH v1 3/4] net/mlx5: support yellow meter action for hierarchy tag rule Shun Hao
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Shun Hao @ 2022-05-13 7:33 UTC (permalink / raw)
To: viacheslavo, matan, orika; +Cc: dev, rasland
This patch adds the support of meter action for yellow meter policy
flow, so can use meter action for both green and yellow policy flows
in meter hierarchy.
Currently must use the same meter within one meter policy. Packets
passing green/yellow policy flow will have previous meter color of
green/yellow in subsequent meter.
Signed-off-by: Shun Hao <shunh@nvidia.com>
Acked-by: Matan Azard <matan@nvidia.com>
---
drivers/net/mlx5/mlx5.c | 8 ++-
drivers/net/mlx5/mlx5.h | 8 ++-
drivers/net/mlx5/mlx5_flow_dv.c | 100 ++++++++++++++++++++++-------
drivers/net/mlx5/mlx5_flow_meter.c | 57 +++++++++++-----
4 files changed, 132 insertions(+), 41 deletions(-)
diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c
index 72b1e35673..f0988712df 100644
--- a/drivers/net/mlx5/mlx5.c
+++ b/drivers/net/mlx5/mlx5.c
@@ -686,10 +686,14 @@ mlx5_aso_flow_mtrs_mng_close(struct mlx5_dev_ctx_shared *sh)
#ifdef HAVE_MLX5_DR_CREATE_ACTION_ASO
for (i = 0; i < MLX5_ASO_MTRS_PER_POOL; i++) {
aso_mtr = &mtr_pool->mtrs[i];
- if (aso_mtr->fm.meter_action)
+ if (aso_mtr->fm.meter_action_g)
claim_zero
(mlx5_glue->destroy_flow_action
- (aso_mtr->fm.meter_action));
+ (aso_mtr->fm.meter_action_g));
+ if (aso_mtr->fm.meter_action_y)
+ claim_zero
+ (mlx5_glue->destroy_flow_action
+ (aso_mtr->fm.meter_action_y));
}
#endif /* HAVE_MLX5_DR_CREATE_ACTION_ASO */
claim_zero(mlx5_devx_cmd_destroy
diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
index c92ce6b527..ef755ee8cf 100644
--- a/drivers/net/mlx5/mlx5.h
+++ b/drivers/net/mlx5/mlx5.h
@@ -869,8 +869,10 @@ struct mlx5_flow_meter_info {
/**< Use count. */
struct mlx5_indexed_pool *flow_ipool;
/**< Index pool for flow id. */
- void *meter_action;
+ void *meter_action_g;
/**< Flow meter action. */
+ void *meter_action_y;
+ /**< Flow meter action for yellow init_color. */
};
/* PPS(packets per second) map to BPS(Bytes per second).
@@ -1873,6 +1875,10 @@ struct mlx5_flow_meter_policy *mlx5_flow_meter_policy_find
(struct rte_eth_dev *dev,
uint32_t policy_id,
uint32_t *policy_idx);
+struct mlx5_flow_meter_info *
+mlx5_flow_meter_hierarchy_next_meter(struct mlx5_priv *priv,
+ struct mlx5_flow_meter_policy *policy,
+ uint32_t *mtr_idx);
struct mlx5_flow_meter_policy *
mlx5_flow_meter_hierarchy_get_final_policy(struct rte_eth_dev *dev,
struct mlx5_flow_meter_policy *policy);
diff --git a/drivers/net/mlx5/mlx5_flow_dv.c b/drivers/net/mlx5/mlx5_flow_dv.c
index f9c56204c4..ee7a32acc8 100644
--- a/drivers/net/mlx5/mlx5_flow_dv.c
+++ b/drivers/net/mlx5/mlx5_flow_dv.c
@@ -6534,13 +6534,13 @@ flow_dv_mtr_alloc(struct rte_eth_dev *dev)
struct mlx5_aso_mtr_pool,
mtrs[mtr_free->offset]);
mtr_idx = MLX5_MAKE_MTR_IDX(pool->index, mtr_free->offset);
- if (!mtr_free->fm.meter_action) {
+ if (!mtr_free->fm.meter_action_g) {
#ifdef HAVE_MLX5_DR_CREATE_ACTION_ASO
struct rte_flow_error error;
uint8_t reg_id;
reg_id = mlx5_flow_get_reg_id(dev, MLX5_MTR_COLOR, 0, &error);
- mtr_free->fm.meter_action =
+ mtr_free->fm.meter_action_g =
mlx5_glue->dv_create_flow_action_aso
(priv->sh->rx_domain,
pool->devx_obj->obj,
@@ -6548,7 +6548,7 @@ flow_dv_mtr_alloc(struct rte_eth_dev *dev)
(1 << MLX5_FLOW_COLOR_GREEN),
reg_id - REG_C_0);
#endif /* HAVE_MLX5_DR_CREATE_ACTION_ASO */
- if (!mtr_free->fm.meter_action) {
+ if (!mtr_free->fm.meter_action_g) {
flow_dv_aso_mtr_release_to_pool(dev, mtr_idx);
return 0;
}
@@ -13401,7 +13401,7 @@ flow_dv_translate(struct rte_eth_dev *dev,
NULL, "Failed to get meter in flow.");
/* Set the meter action. */
dev_flow->dv.actions[actions_n++] =
- wks->fm->meter_action;
+ wks->fm->meter_action_g;
action_flags |= MLX5_FLOW_ACTION_METER;
break;
case RTE_FLOW_ACTION_TYPE_SET_IPV4_DSCP:
@@ -15427,7 +15427,7 @@ __flow_dv_destroy_sub_policy_rules(struct rte_eth_dev *dev,
for (i = 0; i < RTE_COLORS; i++) {
next_fm = NULL;
- if (i == RTE_COLOR_GREEN && policy &&
+ if (i <= RTE_COLOR_YELLOW && policy &&
policy->act_cnt[i].fate_action == MLX5_FLOW_FATE_MTR)
next_fm = mlx5_flow_meter_find(priv,
policy->act_cnt[i].next_mtr_id, NULL);
@@ -15551,6 +15551,51 @@ flow_dv_destroy_mtr_policy_acts(struct rte_eth_dev *dev,
mtr_policy->dr_drop_action[j] = NULL;
}
+/**
+ * Create yellow action for color aware meter.
+ *
+ * @param[in] dev
+ * Pointer to the Ethernet device structure.
+ * @param[in] fm
+ * Meter information table.
+ * @param[out] error
+ * Perform verbose error reporting if not NULL. Initialized in case of
+ * error only.
+ *
+ * @return
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+__flow_dv_create_mtr_yellow_action(struct rte_eth_dev *dev,
+ struct mlx5_flow_meter_info *fm,
+ struct rte_mtr_error *error)
+{
+#ifdef HAVE_MLX5_DR_CREATE_ACTION_ASO
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct rte_flow_error flow_err;
+ struct mlx5_aso_mtr *aso_mtr;
+ struct mlx5_aso_mtr_pool *pool;
+ uint8_t reg_id;
+
+ aso_mtr = container_of(fm, struct mlx5_aso_mtr, fm);
+ pool = container_of(aso_mtr, struct mlx5_aso_mtr_pool, mtrs[aso_mtr->offset]);
+ reg_id = mlx5_flow_get_reg_id(dev, MLX5_MTR_COLOR, 0, &flow_err);
+ fm->meter_action_y =
+ mlx5_glue->dv_create_flow_action_aso(priv->sh->rx_domain,
+ pool->devx_obj->obj,
+ aso_mtr->offset,
+ (1 << MLX5_FLOW_COLOR_YELLOW),
+ reg_id - REG_C_0);
+#else
+ RTE_SET_USED(dev);
+#endif
+ if (!fm->meter_action_y) {
+ return -rte_mtr_error_set(error, EINVAL, RTE_MTR_ERROR_TYPE_MTR_ID, NULL,
+ "Fail to create yellow meter action.");
+ }
+ return 0;
+}
+
/**
* Create policy action per domain, lock free,
* (mutex should be acquired by caller).
@@ -15870,7 +15915,7 @@ __flow_dv_create_domain_policy_acts(struct rte_eth_dev *dev,
break;
}
/*
- * No need to check meter hierarchy for Y or R colors
+ * No need to check meter hierarchy for R colors
* here since it is done in the validation stage.
*/
case RTE_FLOW_ACTION_TYPE_METER:
@@ -15921,6 +15966,10 @@ __flow_dv_create_domain_policy_acts(struct rte_eth_dev *dev,
action_flags |=
MLX5_FLOW_ACTION_SET_TAG;
}
+ if (i == RTE_COLOR_YELLOW && next_fm->color_aware &&
+ !next_fm->meter_action_y)
+ if (__flow_dv_create_mtr_yellow_action(dev, next_fm, error))
+ return -rte_errno;
act_cnt->fate_action = MLX5_FLOW_FATE_MTR;
act_cnt->next_mtr_id = next_fm->meter_id;
act_cnt->next_sub_policy = NULL;
@@ -16519,7 +16568,7 @@ __flow_dv_create_policy_acts_rules(struct rte_eth_dev *dev,
struct mlx5_flow_dv_tag_resource *tag;
struct mlx5_flow_dv_port_id_action_resource *port_action;
struct mlx5_hrxq *hrxq;
- struct mlx5_flow_meter_info *next_fm = NULL;
+ struct mlx5_flow_meter_info *next_fm[RTE_COLORS] = {NULL};
struct mlx5_flow_meter_policy *next_policy;
struct mlx5_flow_meter_sub_policy *next_sub_policy;
struct mlx5_flow_tbl_data_entry *tbl_data;
@@ -16540,30 +16589,31 @@ __flow_dv_create_policy_acts_rules(struct rte_eth_dev *dev,
acts[i].actions_n = 1;
continue;
}
- if (i == RTE_COLOR_GREEN &&
- mtr_policy->act_cnt[i].fate_action == MLX5_FLOW_FATE_MTR) {
+ if (mtr_policy->act_cnt[i].fate_action == MLX5_FLOW_FATE_MTR) {
struct rte_flow_attr attr = {
.transfer = transfer
};
- next_fm = mlx5_flow_meter_find(priv,
+ next_fm[i] = mlx5_flow_meter_find(priv,
mtr_policy->act_cnt[i].next_mtr_id,
NULL);
- if (!next_fm) {
+ if (!next_fm[i]) {
DRV_LOG(ERR,
"Failed to get next hierarchy meter.");
goto err_exit;
}
- if (mlx5_flow_meter_attach(priv, next_fm,
+ if (mlx5_flow_meter_attach(priv, next_fm[i],
&attr, &error)) {
DRV_LOG(ERR, "%s", error.message);
- next_fm = NULL;
+ next_fm[i] = NULL;
goto err_exit;
}
/* Meter action must be the first for TX. */
if (mtr_first) {
acts[i].dv_actions[acts[i].actions_n] =
- next_fm->meter_action;
+ (next_fm[i]->color_aware && i == RTE_COLOR_YELLOW) ?
+ next_fm[i]->meter_action_y :
+ next_fm[i]->meter_action_g;
acts[i].actions_n++;
}
}
@@ -16621,14 +16671,16 @@ __flow_dv_create_policy_acts_rules(struct rte_eth_dev *dev,
acts[i].actions_n++;
break;
case MLX5_FLOW_FATE_MTR:
- if (!next_fm) {
+ if (!next_fm[i]) {
DRV_LOG(ERR,
"No next hierarchy meter.");
goto err_exit;
}
if (!mtr_first) {
acts[i].dv_actions[acts[i].actions_n] =
- next_fm->meter_action;
+ (next_fm[i]->color_aware && i == RTE_COLOR_YELLOW) ?
+ next_fm[i]->meter_action_y :
+ next_fm[i]->meter_action_g;
acts[i].actions_n++;
}
if (mtr_policy->act_cnt[i].next_sub_policy) {
@@ -16637,7 +16689,7 @@ __flow_dv_create_policy_acts_rules(struct rte_eth_dev *dev,
} else {
next_policy =
mlx5_flow_meter_policy_find(dev,
- next_fm->policy_id, NULL);
+ next_fm[i]->policy_id, NULL);
MLX5_ASSERT(next_policy);
next_sub_policy =
next_policy->sub_policys[domain][0];
@@ -16664,8 +16716,9 @@ __flow_dv_create_policy_acts_rules(struct rte_eth_dev *dev,
}
return 0;
err_exit:
- if (next_fm)
- mlx5_flow_meter_detach(priv, next_fm);
+ for (i = 0; i < RTE_COLORS; i++)
+ if (next_fm[i])
+ mlx5_flow_meter_detach(priv, next_fm[i]);
return -1;
}
@@ -17181,8 +17234,9 @@ flow_dv_meter_sub_policy_rss_prepare(struct rte_eth_dev *dev,
DRV_LOG(ERR, "Exceed max meter number in hierarchy.");
return NULL;
}
- next_fm = mlx5_flow_meter_find(priv,
- mtr_policy->act_cnt[RTE_COLOR_GREEN].next_mtr_id, NULL);
+ rte_spinlock_lock(&mtr_policy->sl);
+ next_fm = mlx5_flow_meter_hierarchy_next_meter(priv, mtr_policy, NULL);
+ rte_spinlock_unlock(&mtr_policy->sl);
if (!next_fm) {
DRV_LOG(ERR, "Failed to get next meter in hierarchy.");
return NULL;
@@ -17330,11 +17384,11 @@ flow_dv_meter_hierarchy_rule_create(struct rte_eth_dev *dev,
struct mlx5_flow_tbl_data_entry, tbl);
act_cnt = &mtr_policy->act_cnt[i];
if (mtr_first) {
- acts.dv_actions[0] = next_fm->meter_action;
+ acts.dv_actions[0] = next_fm->meter_action_g;
acts.dv_actions[1] = act_cnt->modify_hdr->action;
} else {
acts.dv_actions[0] = act_cnt->modify_hdr->action;
- acts.dv_actions[1] = next_fm->meter_action;
+ acts.dv_actions[1] = next_fm->meter_action_g;
}
acts.dv_actions[2] = tbl_data->jump.action;
acts.actions_n = 3;
diff --git a/drivers/net/mlx5/mlx5_flow_meter.c b/drivers/net/mlx5/mlx5_flow_meter.c
index b52de70afa..22f6ca7faf 100644
--- a/drivers/net/mlx5/mlx5_flow_meter.c
+++ b/drivers/net/mlx5/mlx5_flow_meter.c
@@ -611,6 +611,36 @@ mlx5_flow_meter_policy_find(struct rte_eth_dev *dev,
return NULL;
}
+/**
+ * Get the next meter from one meter's policy in hierarchy chain.
+ * Lock free, mutex should be acquired by caller.
+ *
+ * @param[in] priv
+ * Pointer to mlx5_priv.
+ * @param[in] policy
+ * Pointer to flow meter policy.
+ * @param[out] mtr_idx
+ * Pointer to Meter index.
+ *
+ * @return
+ * Pointer to the next meter, or NULL when fail.
+ */
+struct mlx5_flow_meter_info *
+mlx5_flow_meter_hierarchy_next_meter(struct mlx5_priv *priv,
+ struct mlx5_flow_meter_policy *policy,
+ uint32_t *mtr_idx)
+{
+ int i;
+
+ for (i = 0; i < MLX5_MTR_RTE_COLORS; i++) {
+ if (policy->act_cnt[i].fate_action == MLX5_FLOW_FATE_MTR)
+ return mlx5_flow_meter_find(priv,
+ policy->act_cnt[i].next_mtr_id,
+ mtr_idx);
+ }
+ return NULL;
+}
+
/**
* Get the last meter's policy from one meter's policy in hierarchy.
*
@@ -631,8 +661,9 @@ mlx5_flow_meter_hierarchy_get_final_policy(struct rte_eth_dev *dev,
struct mlx5_flow_meter_policy *next_policy = policy;
while (next_policy->is_hierarchy) {
- next_fm = mlx5_flow_meter_find(priv,
- next_policy->act_cnt[RTE_COLOR_GREEN].next_mtr_id, NULL);
+ rte_spinlock_lock(&next_policy->sl);
+ next_fm = mlx5_flow_meter_hierarchy_next_meter(priv, next_policy, NULL);
+ rte_spinlock_unlock(&next_policy->sl);
if (!next_fm || next_fm->def_policy)
return NULL;
next_policy = mlx5_flow_meter_policy_find(dev,
@@ -1105,9 +1136,9 @@ mlx5_flow_meter_action_modify(struct mlx5_priv *priv,
ebs_mantissa, val);
}
/* Apply modifications to meter only if it was created. */
- if (fm->meter_action) {
+ if (fm->meter_action_g) {
ret = mlx5_glue->dv_modify_flow_action_meter
- (fm->meter_action, &mod_attr,
+ (fm->meter_action_g, &mod_attr,
rte_cpu_to_be_64(modify_bits));
if (ret)
return ret;
@@ -1908,7 +1939,7 @@ mlx5_flow_meter_attach(struct mlx5_priv *priv,
rte_spinlock_unlock(&fm->sl);
} else {
rte_spinlock_lock(&fm->sl);
- if (fm->meter_action) {
+ if (fm->meter_action_g) {
if (fm->shared &&
attr->transfer == fm->transfer &&
attr->ingress == fm->ingress &&
@@ -1928,9 +1959,9 @@ mlx5_flow_meter_attach(struct mlx5_priv *priv,
fm->transfer = attr->transfer;
fm->ref_cnt = 1;
/* This also creates the meter object. */
- fm->meter_action = mlx5_flow_meter_action_create(priv,
+ fm->meter_action_g = mlx5_flow_meter_action_create(priv,
fm);
- if (!fm->meter_action) {
+ if (!fm->meter_action_g) {
fm->ref_cnt = 0;
fm->ingress = 0;
fm->egress = 0;
@@ -1962,8 +1993,8 @@ mlx5_flow_meter_detach(struct mlx5_priv *priv,
rte_spinlock_lock(&fm->sl);
MLX5_ASSERT(fm->ref_cnt);
if (--fm->ref_cnt == 0 && !priv->sh->meter_aso_en) {
- mlx5_glue->destroy_flow_action(fm->meter_action);
- fm->meter_action = NULL;
+ mlx5_glue->destroy_flow_action(fm->meter_action_g);
+ fm->meter_action_g = NULL;
fm->ingress = 0;
fm->egress = 0;
fm->transfer = 0;
@@ -2040,9 +2071,7 @@ mlx5_flow_meter_flush_hierarchy(struct rte_eth_dev *dev,
MLX5_ASSERT(policy);
while (!fm->ref_cnt && policy->is_hierarchy) {
policy_id = fm->policy_id;
- next_fm = mlx5_flow_meter_find(priv,
- policy->act_cnt[RTE_COLOR_GREEN].next_mtr_id,
- &next_mtr_idx);
+ next_fm = mlx5_flow_meter_hierarchy_next_meter(priv, policy, &next_mtr_idx);
if (next_fm) {
next_policy = mlx5_flow_meter_policy_find(dev,
next_fm->policy_id,
@@ -2120,9 +2149,7 @@ mlx5_flow_meter_flush_all_hierarchies(struct rte_eth_dev *dev,
policy = sub_policy->main_policy;
if (!policy || !policy->is_hierarchy || policy->ref_cnt)
continue;
- next_fm = mlx5_flow_meter_find(priv,
- policy->act_cnt[RTE_COLOR_GREEN].next_mtr_id,
- &mtr_idx);
+ next_fm = mlx5_flow_meter_hierarchy_next_meter(priv, policy, &mtr_idx);
if (__mlx5_flow_meter_policy_delete(dev, i, policy,
error, true))
return -rte_mtr_error_set(error,
--
2.20.0
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH v1 3/4] net/mlx5: support yellow meter action for hierarchy tag rule
2022-05-13 7:33 [PATCH v1 0/4] Enable yellow meter hierarchy Shun Hao
2022-05-13 7:33 ` [PATCH v1 1/4] net/mlx5: support previous meter color aware Shun Hao
2022-05-13 7:33 ` [PATCH v1 2/4] net/mlx5: support yellow meter action in hierarchy Shun Hao
@ 2022-05-13 7:33 ` Shun Hao
2022-05-13 7:33 ` [PATCH v1 4/4] net/mlx5: add validation for yellow meter action Shun Hao
2022-05-16 12:59 ` [PATCH v1 0/4] Enable yellow meter hierarchy Raslan Darawsheh
4 siblings, 0 replies; 6+ messages in thread
From: Shun Hao @ 2022-05-13 7:33 UTC (permalink / raw)
To: viacheslavo, matan, orika; +Cc: dev, rasland
When a hierarchy meter is shared by other ports, it's needed to iterate
all meter policies in hierarchy to create tag rules, to set packet with
next meter ID, which will be used by related meter drop count.
This patch adds the tag rule for yellow support in hierarchy, so both
green/yellow policy flows can set the correct meter ID.
Signed-off-by: Shun Hao <shunh@nvidia.com>
Acked-by: Matan Azard <matan@nvidia.com>
---
drivers/net/mlx5/mlx5_flow_dv.c | 270 +++++++++++++++++++++-----------
1 file changed, 175 insertions(+), 95 deletions(-)
diff --git a/drivers/net/mlx5/mlx5_flow_dv.c b/drivers/net/mlx5/mlx5_flow_dv.c
index ee7a32acc8..287095cceb 100644
--- a/drivers/net/mlx5/mlx5_flow_dv.c
+++ b/drivers/net/mlx5/mlx5_flow_dv.c
@@ -17293,6 +17293,68 @@ flow_dv_meter_sub_policy_rss_prepare(struct rte_eth_dev *dev,
return NULL;
}
+/**
+ * Check if need to create hierarchy tag rule.
+ *
+ * @param[in] priv
+ * Pointer to mlx5_priv.
+ * @param[in] mtr_policy
+ * Pointer to current meter policy.
+ * @param[in] src_port
+ * The src port this extra rule should use.
+ * @param[out] next_fm
+ * Pointer to next meter in hierarchy.
+ * @param[out] skip
+ * Indicate if skip the tag rule creation.
+ * @param[out] error
+ * Perform verbose error reporting if not NULL.
+ * @return
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_meter_hierarchy_skip_tag_rule(struct mlx5_priv *priv,
+ struct mlx5_flow_meter_policy *mtr_policy,
+ int32_t src_port,
+ struct mlx5_flow_meter_info **next_fm,
+ bool *skip,
+ struct rte_flow_error *error)
+{
+ struct mlx5_flow_meter_sub_policy *sub_policy;
+ struct mlx5_sub_policy_color_rule *color_rule;
+ uint32_t domain = MLX5_MTR_DOMAIN_TRANSFER;
+ int ret = 0;
+ int i;
+
+ *next_fm = NULL;
+ *skip = false;
+ rte_spinlock_lock(&mtr_policy->sl);
+ if (!mtr_policy->is_hierarchy)
+ goto exit;
+ *next_fm = mlx5_flow_meter_hierarchy_next_meter(priv, mtr_policy, NULL);
+ if (!*next_fm) {
+ ret = rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ACTION,
+ NULL, "Failed to find next meter in hierarchy.");
+ goto exit;
+ }
+ if (!(*next_fm)->drop_cnt) {
+ *skip = true;
+ goto exit;
+ }
+ sub_policy = mtr_policy->sub_policys[domain][0];
+ for (i = 0; i < MLX5_MTR_RTE_COLORS; i++) {
+ if (mtr_policy->act_cnt[i].fate_action != MLX5_FLOW_FATE_MTR)
+ continue;
+ TAILQ_FOREACH(color_rule, &sub_policy->color_rules[i], next_port)
+ if (color_rule->src_port == src_port) {
+ *skip = true;
+ goto exit;
+ }
+ }
+exit:
+ rte_spinlock_unlock(&mtr_policy->sl);
+ return ret;
+}
+
/**
* Create the sub policy tag rule for all meters in hierarchy.
*
@@ -17336,111 +17398,129 @@ flow_dv_meter_hierarchy_rule_create(struct rte_eth_dev *dev,
.reserved = 0,
};
uint32_t domain = MLX5_MTR_DOMAIN_TRANSFER;
- int i;
+ struct {
+ struct mlx5_flow_meter_policy *fm_policy;
+ struct mlx5_flow_meter_info *next_fm;
+ struct mlx5_sub_policy_color_rule *tag_rule[MLX5_MTR_RTE_COLORS];
+ } fm_info[MLX5_MTR_CHAIN_MAX_NUM] = { {0} };
+ uint32_t fm_cnt = 0;
+ uint32_t i, j;
- mtr_policy = mlx5_flow_meter_policy_find(dev, fm->policy_id, NULL);
- MLX5_ASSERT(mtr_policy);
- if (!mtr_policy->is_hierarchy)
- return 0;
- next_fm = mlx5_flow_meter_find(priv,
- mtr_policy->act_cnt[RTE_COLOR_GREEN].next_mtr_id, NULL);
- if (!next_fm) {
- return rte_flow_error_set(error, EINVAL,
- RTE_FLOW_ERROR_TYPE_ACTION, NULL,
- "Failed to find next meter in hierarchy.");
- }
- if (!next_fm->drop_cnt)
- goto exit;
color_reg_c_idx = mlx5_flow_get_reg_id(dev, MLX5_MTR_COLOR, 0, error);
- sub_policy = mtr_policy->sub_policys[domain][0];
- for (i = 0; i < RTE_COLORS; i++) {
- bool rule_exist = false;
- struct mlx5_meter_policy_action_container *act_cnt;
+ /* Get all fms who need to create the tag color rule. */
+ do {
+ bool skip = false;
- if (i >= RTE_COLOR_YELLOW)
- break;
- TAILQ_FOREACH(color_rule,
- &sub_policy->color_rules[i], next_port)
- if (color_rule->src_port == src_port) {
- rule_exist = true;
- break;
- }
- if (rule_exist)
- continue;
- color_rule = mlx5_malloc(MLX5_MEM_ZERO,
- sizeof(struct mlx5_sub_policy_color_rule),
- 0, SOCKET_ID_ANY);
- if (!color_rule)
- return rte_flow_error_set(error, ENOMEM,
- RTE_FLOW_ERROR_TYPE_ACTION,
- NULL, "No memory to create tag color rule.");
- color_rule->src_port = src_port;
- attr.priority = i;
- next_policy = mlx5_flow_meter_policy_find(dev,
- next_fm->policy_id, NULL);
- MLX5_ASSERT(next_policy);
- next_sub_policy = next_policy->sub_policys[domain][0];
- tbl_data = container_of(next_sub_policy->tbl_rsc,
- struct mlx5_flow_tbl_data_entry, tbl);
- act_cnt = &mtr_policy->act_cnt[i];
- if (mtr_first) {
- acts.dv_actions[0] = next_fm->meter_action_g;
- acts.dv_actions[1] = act_cnt->modify_hdr->action;
- } else {
- acts.dv_actions[0] = act_cnt->modify_hdr->action;
- acts.dv_actions[1] = next_fm->meter_action_g;
- }
- acts.dv_actions[2] = tbl_data->jump.action;
- acts.actions_n = 3;
- if (mlx5_flow_meter_attach(priv, next_fm, &attr, error)) {
- next_fm = NULL;
- goto err_exit;
- }
- if (__flow_dv_create_policy_matcher(dev, color_reg_c_idx,
- MLX5_MTR_POLICY_MATCHER_PRIO, sub_policy,
- &attr, true, item,
- &color_rule->matcher, error)) {
- rte_flow_error_set(error, errno,
- RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
- "Failed to create hierarchy meter matcher.");
+ mtr_policy = mlx5_flow_meter_policy_find(dev, fm->policy_id, NULL);
+ MLX5_ASSERT(mtr_policy);
+ if (mlx5_meter_hierarchy_skip_tag_rule(priv, mtr_policy, src_port,
+ &next_fm, &skip, error))
goto err_exit;
+ if (next_fm && !skip) {
+ fm_info[fm_cnt].fm_policy = mtr_policy;
+ fm_info[fm_cnt].next_fm = next_fm;
+ if (++fm_cnt >= MLX5_MTR_CHAIN_MAX_NUM) {
+ rte_flow_error_set(error, errno,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+ "Exceed max meter number in hierarchy.");
+ goto err_exit;
+ }
}
- if (__flow_dv_create_policy_flow(dev, color_reg_c_idx,
- (enum rte_color)i,
- color_rule->matcher->matcher_object,
- acts.actions_n, acts.dv_actions,
- true, item,
- &color_rule->rule, &attr)) {
- rte_flow_error_set(error, errno,
- RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
- "Failed to create hierarchy meter rule.");
- goto err_exit;
+ fm = next_fm;
+ } while (fm);
+ /* Create tag color rules for all needed fms. */
+ for (i = 0; i < fm_cnt; i++) {
+ void *mtr_action;
+
+ mtr_policy = fm_info[i].fm_policy;
+ rte_spinlock_lock(&mtr_policy->sl);
+ sub_policy = mtr_policy->sub_policys[domain][0];
+ for (j = 0; j < MLX5_MTR_RTE_COLORS; j++) {
+ if (mtr_policy->act_cnt[j].fate_action != MLX5_FLOW_FATE_MTR)
+ continue;
+ color_rule = mlx5_malloc(MLX5_MEM_ZERO,
+ sizeof(struct mlx5_sub_policy_color_rule),
+ 0, SOCKET_ID_ANY);
+ if (!color_rule) {
+ rte_spinlock_unlock(&mtr_policy->sl);
+ rte_flow_error_set(error, ENOMEM,
+ RTE_FLOW_ERROR_TYPE_ACTION, NULL,
+ "No memory to create tag color rule.");
+ goto err_exit;
+ }
+ color_rule->src_port = src_port;
+ next_fm = fm_info[i].next_fm;
+ if (mlx5_flow_meter_attach(priv, next_fm, &attr, error)) {
+ mlx5_free(color_rule);
+ rte_spinlock_unlock(&mtr_policy->sl);
+ goto err_exit;
+ }
+ fm_info[i].tag_rule[j] = color_rule;
+ TAILQ_INSERT_TAIL(&sub_policy->color_rules[j], color_rule, next_port);
+ /* Prepare to create color rule. */
+ mtr_action = (next_fm->color_aware && j == RTE_COLOR_YELLOW) ?
+ next_fm->meter_action_y :
+ next_fm->meter_action_g;
+ next_policy = mlx5_flow_meter_policy_find(dev, next_fm->policy_id, NULL);
+ MLX5_ASSERT(next_policy);
+ next_sub_policy = next_policy->sub_policys[domain][0];
+ tbl_data = container_of(next_sub_policy->tbl_rsc,
+ struct mlx5_flow_tbl_data_entry, tbl);
+ if (mtr_first) {
+ acts.dv_actions[0] = mtr_action;
+ acts.dv_actions[1] = mtr_policy->act_cnt[j].modify_hdr->action;
+ } else {
+ acts.dv_actions[0] = mtr_policy->act_cnt[j].modify_hdr->action;
+ acts.dv_actions[1] = mtr_action;
+ }
+ acts.dv_actions[2] = tbl_data->jump.action;
+ acts.actions_n = 3;
+ if (__flow_dv_create_policy_matcher(dev, color_reg_c_idx,
+ MLX5_MTR_POLICY_MATCHER_PRIO, sub_policy,
+ &attr, true, item, &color_rule->matcher, error)) {
+ rte_spinlock_unlock(&mtr_policy->sl);
+ rte_flow_error_set(error, errno,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+ "Failed to create hierarchy meter matcher.");
+ goto err_exit;
+ }
+ if (__flow_dv_create_policy_flow(dev, color_reg_c_idx, (enum rte_color)j,
+ color_rule->matcher->matcher_object,
+ acts.actions_n, acts.dv_actions,
+ true, item, &color_rule->rule, &attr)) {
+ rte_spinlock_unlock(&mtr_policy->sl);
+ rte_flow_error_set(error, errno,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+ "Failed to create hierarchy meter rule.");
+ goto err_exit;
+ }
}
- TAILQ_INSERT_TAIL(&sub_policy->color_rules[i],
- color_rule, next_port);
+ rte_spinlock_unlock(&mtr_policy->sl);
}
-exit:
- /**
- * Recursive call to iterate all meters in hierarchy and
- * create needed rules.
- */
- return flow_dv_meter_hierarchy_rule_create(dev, next_fm,
- src_port, item, error);
+ return 0;
err_exit:
- if (color_rule) {
- if (color_rule->rule)
- mlx5_flow_os_destroy_flow(color_rule->rule);
- if (color_rule->matcher) {
- struct mlx5_flow_tbl_data_entry *tbl =
- container_of(color_rule->matcher->tbl,
- typeof(*tbl), tbl);
- mlx5_list_unregister(tbl->matchers,
- &color_rule->matcher->entry);
+ for (i = 0; i < fm_cnt; i++) {
+ mtr_policy = fm_info[i].fm_policy;
+ rte_spinlock_lock(&mtr_policy->sl);
+ sub_policy = mtr_policy->sub_policys[domain][0];
+ for (j = 0; j < MLX5_MTR_RTE_COLORS; j++) {
+ color_rule = fm_info[i].tag_rule[j];
+ if (!color_rule)
+ continue;
+ if (color_rule->rule)
+ mlx5_flow_os_destroy_flow(color_rule->rule);
+ if (color_rule->matcher) {
+ struct mlx5_flow_tbl_data_entry *tbl =
+ container_of(color_rule->matcher->tbl, typeof(*tbl), tbl);
+ mlx5_list_unregister(tbl->matchers, &color_rule->matcher->entry);
+ }
+ if (fm_info[i].next_fm)
+ mlx5_flow_meter_detach(priv, fm_info[i].next_fm);
+ TAILQ_REMOVE(&sub_policy->color_rules[j], color_rule, next_port);
+ mlx5_free(color_rule);
}
- mlx5_free(color_rule);
+ rte_spinlock_unlock(&mtr_policy->sl);
}
- if (next_fm)
- mlx5_flow_meter_detach(priv, next_fm);
return -rte_errno;
}
--
2.20.0
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH v1 4/4] net/mlx5: add validation for yellow meter action
2022-05-13 7:33 [PATCH v1 0/4] Enable yellow meter hierarchy Shun Hao
` (2 preceding siblings ...)
2022-05-13 7:33 ` [PATCH v1 3/4] net/mlx5: support yellow meter action for hierarchy tag rule Shun Hao
@ 2022-05-13 7:33 ` Shun Hao
2022-05-16 12:59 ` [PATCH v1 0/4] Enable yellow meter hierarchy Raslan Darawsheh
4 siblings, 0 replies; 6+ messages in thread
From: Shun Hao @ 2022-05-13 7:33 UTC (permalink / raw)
To: viacheslavo, matan, orika; +Cc: dev, rasland
Yellow meter action support is added in meter hierarchy validation.
If one color uses meter action, the other can only use NULL action
or the same meter action. And only shared meter is supported.
Signed-off-by: Shun Hao <shunh@nvidia.com>
Acked-by: Matan Azard <matan@nvidia.com>
---
doc/guides/nics/mlx5.rst | 6 +++--
drivers/net/mlx5/mlx5_flow_dv.c | 40 +++++++++++++++++----------------
2 files changed, 25 insertions(+), 21 deletions(-)
diff --git a/doc/guides/nics/mlx5.rst b/doc/guides/nics/mlx5.rst
index 4805d08a76..a47a3fc576 100644
--- a/doc/guides/nics/mlx5.rst
+++ b/doc/guides/nics/mlx5.rst
@@ -443,13 +443,15 @@ Limitations
- yellow: NULL or END.
- RED: DROP / END.
- The only supported meter policy actions:
- - green: QUEUE, RSS, PORT_ID, REPRESENTED_PORT, JUMP, DROP, MARK and SET_TAG.
- - yellow: QUEUE, RSS, PORT_ID, REPRESENTED_PORT, JUMP, DROP, MARK and SET_TAG.
+ - green: QUEUE, RSS, PORT_ID, REPRESENTED_PORT, JUMP, DROP, MARK, METER and SET_TAG.
+ - yellow: QUEUE, RSS, PORT_ID, REPRESENTED_PORT, JUMP, DROP, MARK, METER and SET_TAG.
- RED: must be DROP.
- Policy actions of RSS for green and yellow should have the same configuration except queues.
- Policy with RSS/queue action is not supported when ``dv_xmeta_en`` enabled.
+ - If green action is METER, yellow action must be the same METER action or NULL.
- meter profile packet mode is supported.
- meter profiles of RFC2697, RFC2698 and RFC4115 are supported.
+ - RFC4115 implementation is following MEF, meaning yellow traffic may reclaim unused green bandwidth when green token bucket is full.
- Integrity:
diff --git a/drivers/net/mlx5/mlx5_flow_dv.c b/drivers/net/mlx5/mlx5_flow_dv.c
index 287095cceb..9354675e97 100644
--- a/drivers/net/mlx5/mlx5_flow_dv.c
+++ b/drivers/net/mlx5/mlx5_flow_dv.c
@@ -18003,8 +18003,8 @@ flow_dv_validate_policy_mtr_hierarchy(struct rte_eth_dev *dev,
NULL,
"Multiple fate actions not supported.");
*hierarchy_domain = 0;
+ fm = mlx5_flow_meter_find(priv, meter_id, NULL);
while (true) {
- fm = mlx5_flow_meter_find(priv, meter_id, NULL);
if (!fm)
return -rte_mtr_error_set(error, EINVAL,
RTE_MTR_ERROR_TYPE_MTR_ID, NULL,
@@ -18013,6 +18013,10 @@ flow_dv_validate_policy_mtr_hierarchy(struct rte_eth_dev *dev,
return -rte_mtr_error_set(error, EINVAL,
RTE_MTR_ERROR_TYPE_MTR_ID, NULL,
"Non termination meter not supported in hierarchy.");
+ if (!fm->shared)
+ return -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_MTR_ID, NULL,
+ "Only shared meter supported in hierarchy.");
policy = mlx5_flow_meter_policy_find(dev, fm->policy_id, NULL);
MLX5_ASSERT(policy);
/**
@@ -18034,7 +18038,9 @@ flow_dv_validate_policy_mtr_hierarchy(struct rte_eth_dev *dev,
*is_rss = policy->is_rss;
break;
}
- meter_id = policy->act_cnt[RTE_COLOR_GREEN].next_mtr_id;
+ rte_spinlock_lock(&policy->sl);
+ fm = mlx5_flow_meter_hierarchy_next_meter(priv, policy, NULL);
+ rte_spinlock_unlock(&policy->sl);
if (++cnt >= MLX5_MTR_CHAIN_MAX_NUM)
return -rte_mtr_error_set(error, EINVAL,
RTE_MTR_ERROR_TYPE_METER_POLICY, NULL,
@@ -18080,6 +18086,7 @@ flow_dv_validate_mtr_policy_acts(struct rte_eth_dev *dev,
uint8_t def_domain = MLX5_MTR_ALL_DOMAIN_BIT;
uint8_t hierarchy_domain = 0;
const struct rte_flow_action_meter *mtr;
+ const struct rte_flow_action_meter *next_mtr = NULL;
bool def_green = false;
bool def_yellow = false;
const struct rte_flow_action_rss *rss_color[RTE_COLORS] = {NULL};
@@ -18263,25 +18270,12 @@ flow_dv_validate_mtr_policy_acts(struct rte_eth_dev *dev,
++actions_n;
action_flags[i] |= MLX5_FLOW_ACTION_JUMP;
break;
- /*
- * Only the last meter in the hierarchy will support
- * the YELLOW color steering. Then in the meter policy
- * actions list, there should be no other meter inside.
- */
case RTE_FLOW_ACTION_TYPE_METER:
- if (i != RTE_COLOR_GREEN)
- return -rte_mtr_error_set(error,
- ENOTSUP,
- RTE_MTR_ERROR_TYPE_METER_POLICY,
- NULL,
- "Meter hierarchy only supports GREEN color.");
- if (*policy_mode != MLX5_MTR_POLICY_MODE_OG)
- return -rte_mtr_error_set(error,
- ENOTSUP,
- RTE_MTR_ERROR_TYPE_METER_POLICY,
- NULL,
- "No yellow policy should be provided in meter hierarchy.");
mtr = act->conf;
+ if (next_mtr && next_mtr->mtr_id != mtr->mtr_id)
+ return -rte_mtr_error_set(error, ENOTSUP,
+ RTE_MTR_ERROR_TYPE_METER_POLICY, NULL,
+ "Green and Yellow must use the same meter.");
ret = flow_dv_validate_policy_mtr_hierarchy(dev,
mtr->mtr_id,
action_flags[i],
@@ -18293,6 +18287,7 @@ flow_dv_validate_mtr_policy_acts(struct rte_eth_dev *dev,
++actions_n;
action_flags[i] |=
MLX5_FLOW_ACTION_METER_WITH_TERMINATED_POLICY;
+ next_mtr = mtr;
break;
default:
return -rte_mtr_error_set(error, ENOTSUP,
@@ -18378,6 +18373,13 @@ flow_dv_validate_mtr_policy_acts(struct rte_eth_dev *dev,
}
}
}
+ if (next_mtr && *policy_mode == MLX5_MTR_POLICY_MODE_ALL) {
+ if (!(action_flags[RTE_COLOR_GREEN] & action_flags[RTE_COLOR_YELLOW] &
+ MLX5_FLOW_ACTION_METER_WITH_TERMINATED_POLICY))
+ return -rte_mtr_error_set(error, EINVAL, RTE_MTR_ERROR_TYPE_METER_POLICY,
+ NULL,
+ "Meter hierarchy supports meter action only.");
+ }
/* If both colors have RSS, the attributes should be the same. */
if (flow_dv_mtr_policy_rss_compare(rss_color[RTE_COLOR_GREEN],
rss_color[RTE_COLOR_YELLOW]))
--
2.20.0
^ permalink raw reply [flat|nested] 6+ messages in thread
* RE: [PATCH v1 0/4] Enable yellow meter hierarchy
2022-05-13 7:33 [PATCH v1 0/4] Enable yellow meter hierarchy Shun Hao
` (3 preceding siblings ...)
2022-05-13 7:33 ` [PATCH v1 4/4] net/mlx5: add validation for yellow meter action Shun Hao
@ 2022-05-16 12:59 ` Raslan Darawsheh
4 siblings, 0 replies; 6+ messages in thread
From: Raslan Darawsheh @ 2022-05-16 12:59 UTC (permalink / raw)
To: Shun Hao, Slava Ovsiienko, Matan Azrad, Ori Kam; +Cc: dev
Hi,
> -----Original Message-----
> From: Shun Hao <shunh@nvidia.com>
> Sent: Friday, May 13, 2022 10:33 AM
> To: Slava Ovsiienko <viacheslavo@nvidia.com>; Matan Azrad
> <matan@nvidia.com>; Ori Kam <orika@nvidia.com>
> Cc: dev@dpdk.org; Raslan Darawsheh <rasland@nvidia.com>
> Subject: [PATCH v1 0/4] Enable yellow meter hierarchy
>
> To support yellow meter hierarchy, need to support meter action in yellow
> policy flow, so both green and yellow packet can go to next meter. Currently
> there's the limitation that only the same next meter supported for green and
> yellow.
>
> Meanwhile, color aware mode is supported, so the next meter can be aware
> of the previous color, so can have different process strategy.
>
> Shun Hao (4):
> net/mlx5: support previous meter color aware
> net/mlx5: support yellow meter action in hierarchy
> net/mlx5: support yellow meter action for hierarchy tag rule
> net/mlx5: add validation for yellow meter action
>
> doc/guides/nics/mlx5.rst | 6 +-
> drivers/net/mlx5/mlx5.c | 8 +-
> drivers/net/mlx5/mlx5.h | 10 +-
> drivers/net/mlx5/mlx5_flow_aso.c | 19 +-
> drivers/net/mlx5/mlx5_flow_dv.c | 406 +++++++++++++++++++----------
> drivers/net/mlx5/mlx5_flow_meter.c | 60 +++--
> 6 files changed, 345 insertions(+), 164 deletions(-)
>
> --
> 2.20.0
Series applied to next-net-mlx,
Kindest regards,
Raslan Darawsheh
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2022-05-16 13:00 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-05-13 7:33 [PATCH v1 0/4] Enable yellow meter hierarchy Shun Hao
2022-05-13 7:33 ` [PATCH v1 1/4] net/mlx5: support previous meter color aware Shun Hao
2022-05-13 7:33 ` [PATCH v1 2/4] net/mlx5: support yellow meter action in hierarchy Shun Hao
2022-05-13 7:33 ` [PATCH v1 3/4] net/mlx5: support yellow meter action for hierarchy tag rule Shun Hao
2022-05-13 7:33 ` [PATCH v1 4/4] net/mlx5: add validation for yellow meter action Shun Hao
2022-05-16 12:59 ` [PATCH v1 0/4] Enable yellow meter hierarchy Raslan Darawsheh
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).